**The AI Report Generator turns a finished Dashboard into a clear, shareable narrative report, so you can move from "numbers are ready" to "stakeholders understand the story" in seconds instead of hours.**

![AI report generation flow](/images/report_generation.gif)

## Before you start

1. **Build** the Dashboard you want to narrate first: see [Building Dashboards](/docs/building-dashboards).
2. **Confirm** you have access to premium features.
3. **Refine** the Dashboard so it already reflects the analysis you want to communicate (correct filters, date range, and Chart selection).

If the Dashboard is still exploratory, finalise it before generating a report, so the output stays focused and decision-ready.

## Access and credits

Report generation is gated to premium subscriptions and is metered in **query credits**, the same balance used by [AI Chat](/docs/ai-chat-interface).

- **Paid subscriptions**: a monthly credit allowance, with the exact amount depending on your subscription tier.
- **Additional credits**: you can top up at any time by purchasing extra credits.
- **Rollover**: purchased credits **roll over to the next subscription cycle**, so top-ups you do not use this month are not lost.

You can check your current balance and manage purchases in [Settings](/docs/settings).

## Generate a report

1. **Open** the target Dashboard.
2. **Click** the **Report Generation** icon on the top-left of the Dashboard.
3. **Describe** in the prompt field:
   - **Purpose of the report**, for example weekly business review, campaign performance update, monthly operations summary.
   - **What to include**: KPIs, trends, anomalies, comparisons, recommendations, executive summary.
   - **What to exclude**: irrelevant Charts, technical details, specific dimensions, or sections not needed for the audience.
   - **Optional style guidance**: brief vs detailed, technical vs business tone, section order.
4. **Submit** the request.
5. **Wait** a few seconds while OrcaSheets generates the report.

## Prompting tips for better output

- **Be explicit about the audience** (executive, operations, finance, product), so the tone and depth match.
- **Ask for a structure** (summary, key insights, risks, recommendations, next steps), so the report is easy to scan.
- **Mention the date range and comparison window** you care about.
- **Call out constraints**, for example "no speculation", "focus only on conversion and retention", or "exclude regional breakdowns".

The more specific your prompt is, the more useful and on-target the generated report will be.

## Data handling and privacy

During report generation, OrcaSheets may share aggregated and analysed data with the LLM to produce the narrative. **Raw records and PII are not exposed** in this process. This design lets you get a high-quality narrative while preserving data privacy boundaries.

## After generation

Once generated, the report opens directly in OrcaSheets for review. From there you can:

- **Edit and refine** the report content, so the final output matches your voice.
- **Publish** the report for sharing with stakeholders.
- **Download or export** it the same way you would a Dashboard.

This keeps the move from analysis to communication inside the same workflow.

## Related pages

- [Building Dashboards](/docs/building-dashboards)
- [Creating Visualizations](/docs/creating-visualizations)
- [Exporting Data](/docs/exporting-data)
- [AI Chat Interface](/docs/ai-chat-interface)